Akamai v6 - peering issue?

[#1]

I am getting a lot of Facebook images that aren't loading at the moment and they are delivered/hosted by Akamai (address fbcdn-profile-a.akamaihd.net).
A traceroute shows exactly the same issue as with www.java.com (traceroute seems to go no further than PN's network):

Quote

traceroute6: Warning: a2047.dspl.akamai.net has multiple addresses; using 2a02:26f0:5a::173e:3531 traceroute6 to a2047.dspl.akamai.net (2a02:26f0:5a::173e:3531) from 2a02:16c8:2000:1100:ed61:xxxx:xxxx:xxxx, 64 hops max, 12 byte packets 1  2a02:16c8:2000:1100:xxxx:xxxx:xxxx:xxxx  0.376 ms  0.410 ms  0.330 ms 2  2a02:16c8:0:1::16  6.601 ms  6.363 ms  5.997 ms 3  2a02:16c8:1:2::19  4.591 ms  4.706 ms  4.409 ms 4  2a02:16c8::d  4.647 ms  4.575 ms  6.129 ms 5  2a02:16c8:1:2::8  4.629 ms  4.576 ms  4.989 ms 6  * * * 7  * * * 6  * * * 7  * * * 8  * * * 9  * * * 10  * * *

A traceroute from Hurricane's LG works perfectly fine:

Quote

core1.lon1.he.net> traceroute ipv6 2a02:26f0:5a::173e:356a numeric Target 2a02:26f0:5a::173e:356a Hop Packet 1 Packet 2 Packet 3 Hostname 1 10 ms 11 ms <1 ms 10ge3-1.core1.lon2.he.net (2001:470:0:2cd::1) 2 9 ms <1 ms <1 ms ge-2-1-0-zpr1.lnt.cw.net (2001:7f8:4::4f9:1) 3 <1 ms <1 ms <1 ms so-0-1-0-r1-IPv6.de.cw.net (2001:5000:0:1::1) 4 11 ms 2 ms 1 ms akamai-gw2.lns.cw.net (2001:5000:1100:9::2) 5 <1 ms <1 ms <1 ms 2a02:26f0:5a::173e:3531

Quote

core1.lon2.he.net> traceroute ipv6 2a02:26f0:5a::173e:356a numeric Target 2a02:26f0:5a::173e:356a Hop Packet 1 Packet 2 Packet 3 Hostname 1 13 ms <1 ms <1 ms ge-2-1-0-zpr1.lnt.cw.net (2001:7f8:4::4f9:1) 2 1 ms 1 ms 1 ms so-0-1-0-r1-IPv6.de.cw.net (2001:5000:0:1::1) 3 1 ms 1 ms 1 ms akamai-gw2.lns.cw.net (2001:5000:1100:9::2) 4 <1 ms <1 ms <1 ms 2a02:26f0:5a::173e:3531

The strange thing is that if I ping6 akamai-gw2.lns.cw.net (2001:5000:1100:9::2) and then immediately afterwards do a traceroute to a2047.dspl.akamai.net (2a02:26f0:5a::173e:3531), it works fine:

Quote

$ traceroute6 fbcdn-profile-a.akamaihd.net traceroute6: Warning: a2047.dspl.akamai.net has multiple addresses; using 2a02:26f0:5a::173e:3531 traceroute6 to a2047.dspl.akamai.net (2a02:26f0:5a::173e:3531) from 2a02:16c8:2000:1100:xxxx:xxxx:xxxx:xxxx, 64 hops max, 12 byte packets 1  2a02:16c8:2000:1100:xxxx:xxxx:xxxx:xxxx  0.443 ms  0.393 ms  0.336 ms 2  2a02:16c8:0:1::16  128.809 ms  88.936 ms  57.786 ms 3  2a02:16c8:1:2::19  4.883 ms  4.462 ms  4.632 ms 4  2a02:16c8::d  5.034 ms  4.693 ms  4.671 ms 5  2a02:16c8:1:2::8  4.291 ms  4.044 ms  4.420 ms 6  lonap.he.net  4.609 ms  11.987 ms  4.885 ms 7  10ge3-1.core1.lon2.he.net  15.629 ms  4.191 ms  4.699 ms 8  ge-2-1-0-zpr1.lnt.cw.net  4.680 ms  4.639 ms  4.636 ms 9  so-0-1-0-r1-ipv6.de.cw.net  5.188 ms  9.314 ms  7.206 ms 10  akamai-gw2.lns.cw.net  5.511 ms  5.589 ms  5.252 ms 11  2a02:26f0:5a::173e:3531  4.961 ms  5.037 ms  4.885 ms

Any thoughts?

[#2]

Might be something wrong on our side here. We've flagged it over to Paul to have a look.

[#3]

Perfect. Thanks.

[#4]

yeah I had akamai slow loading/timeouts on another site also.

[#5]

I've made some changes to stop traffic going over the Akamai PNI links.  Hopefully this will be a bit more stable while they have a look at what's going on.

[#6]

Working for me fine now - thanks.

[#7]

@paulmh5
Thanks for fixing http://community.plus.net/forum/index.php/topic,126295.0.html as well.

[#8]

Fingers crossed Akamai have sorted the issue on their side now so I've enabled the PNIs again.  Let me know if you see issues with it.

[#9]

Paul,
www.java.com gives "The server at www.java.com is taking too long to respond." over IPv6!

[#10]

Odd, its working ok for me this morning.  Everything else working for you over v6(Facebook etc..)?

[#11]

I can access IPv6 only sites such as http://loopsofzen.co.uk/.
Firefox is accessing some sites (e.g. http://www.internetsociety.org/blog/tech-matters/2014/02/google-ipv6-traffic-passes-3-deployment-rat... via IPv6 (I'm using the 4or6 add-on) - although it's doing it's usual of fast fall back to IPv4 on others.
Tracert:

C:\Users\John>tracert -6 www.java.com
Tracing route to www.java.com [2a02:26f0:26:3:8e00::1ba3]
over a maximum of 30 hops:
  1    <1 ms     1 ms     1 ms  dsldevice.lan [2a02:16c8:<removed>]
  2    21 ms    21 ms    20 ms  2a02:16c8:0:1::17
  3    21 ms    20 ms    20 ms  2a02:16c8:1:2::17
  4    22 ms    20 ms    20 ms  2a02:16c8::c
  5    20 ms    20 ms    20 ms  2a02:16c8:1:2::6
  6    23 ms    24 ms    24 ms  40ge1-3.core1.lon2.he.net [2001:7f8:4::1b1b:1]
  7     *        *        *     Request timed out.
 ...
 19     *        *        *     Request timed out.
 20     *     Destination host unreachable.
Trace complete.

http://ipv6.google.com/ works.
I'm seeing the issue on both pcl and ptw gateways.

[#12]

I can't access http://ipv6.google.com/ at the moment.

[#13]

Both are still working fine for me.
Have you moved to one of the new BNGs?  These will give you a pool address rather than a static if you have that in any fixed config that may cause you a problem.  Failing that I'd suggest checking your MTU size as a few people have had issues with that and some v6 services.

[#14]

I'm not on a bng (does IPSEC go to bng's?).
I've not touched anything to do with MTU - so why would Google suddenly stop working? (I can tracert to it)

[#15]

Anyone else having problems with Facebook images again?
java.com is OK, Facebook just seems a bit iffy from my desktop